Effects of Vitamin DSupplementation on Prevention ofType 2 Diabetes in Patients WithPrediabetes: A Systematic Reviewand Meta-analysisDiabetes Care 2020;43:1650–1658 | https://doi.org/10.2337/dc19-1708

BACKGROUND

While observational studies have shown an association between vitamin D in-sufficiency and diabetes, it is unclear whether intervention with vitamin D supple-ments can lower the risk of type 2 diabetes mellitus (T2DM).

PURPOSE

To assess whether vitamin D supplementation reduces the risk of T2DM in peoplewith prediabetes.

DATA SOURCES

We searchedMEDLINE, Embase, and Cochrane Central Register of Controlled Trials(CENTRAL) from inception to 5 July 2019.

STUDY SELECTION

We included randomized controlled trials assessing vitamin D supplementationversus placebo in relation to new-onset T2DM in people with prediabetes.

DATA EXTRACTION

We screened studies and extracted data from published trials independently.

DATA SYNTHESIS

We identified eight eligible trials with a total of 4,896 subjects. Vitamin D supple-mentationsignificantly reducedtheriskofT2DM(risk ratio [RR]0.89[95%CI0.80–0.99];I25 0%).Benefitwas found innonobesesubjects (RR0.73 [95%CI0.57–0.92]) butnot inobese subjects (RR 0.95 [95% CI 0.84–1.08]) (Pinteraction 5 0.048). The reversion ofprediabetes to normoglycemia occurred in 116 of 548 (21.2%) participants in thevitamin D group and 75 of 532 (14.1%) in the control group. Vitamin D supplemen-tation increased reversion rate of prediabetes to normoglycemia (RR 1.48 [95% CI1.14–1.92]; I2 5 0%.)

LIMITATIONS

Definitionsof prediabetesandnew-onsetdiabetes ineligible studiesweredifferent,and long-term data on outcomes of T2DM prevention were lacking.

CONCLUSIONS

In persons with prediabetes, vitamin D supplementation reduces the risk of T2DMand increases the reversion rate of prediabetes to normoglycemia. The benefitof the prevention of T2DM could be limited to nonobese subjects. Individualparticipant data meta-analyses are needed to confirm these findings.

1West ChinaHospital, SichuanUniversity, Chengdu,Sichuan, China2AffiliatedHospitalofChengduUniversity,Chengdu,Sichuan, China3ChineseUniversity ofHongKong, Shenzhen,Guang-dong, China4Department of Medical Oncology, Thomas Jef-ferson University, Philadelphia, PA5Sidney Kimmel Medical College, Thomas Jeffer-son University, Philadelphia, PA6University of Pittsburgh Medical Center, Uni-versity of Pittsburgh, Pittsburgh, PA7Sichuan University Library, Sichuan University,Chengdu, Sichuan, China

Corresponding author: Fang Fang, fangfang1057@outlook.com

Received 25 August 2019 and accepted 19 Feb-ruary 2020

This article contains supplementary material onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc19-1708/-/DC1.

Y.Z. and H.T. contributed equally to this work.

© 2020 by the American Diabetes Association.Readersmayuse this article as longas thework isproperly cited, the use is educational and not forprofit, and the work is not altered. More infor-mation is availableathttps://www.diabetesjournals.org/content/license.

See accompanying articles, pp. 1375and 1659.

Yu Zhang,1,2 Huiwen Tan,1 Jingjing Tang,3

Jing Li,1 Weelic Chong,4 Yang Hai,5

Yuning Feng,2 L. Dade Lunsford,6 Ping Xu,7

Desheng Jia,1 and Fang Fang1

1650 Diabetes Care Volume 43, July 2020

META-ANALYSIS

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/43/7/1650/630440/dc191708.pdf by guest on 04 January 2022

Observational studies have suggested thatlowserum25-hydroxyvitaminD[25(OH)D]is inversely associatedwith the incidenceof type 2 diabetes mellitus (T2DM) (1–3).The hypothesis that vitamin D status mayinfluence the risk of T2DM is biologicallyplausible because in vivo and in vitrostudies have proposed potential roles ofvitamin D in impaired pancreatic b-cellfunction and insulin sensitivity (4,5). How-ever, evidence from interventional trialsevaluating the efficacy of vitamin D sup-plementation for diabetes preventionat a population level has been incon-clusive. Published systematic reviews andmeta-analyses have suggested that vita-min D supplementation does not lowerthe risk of developing T2DM among peo-ple with prediabetes (6–10). They in-cludedmostly small trials and thus wereinsufficiently powered. Some reviewsincluded mixed interventions (e.g., Vi-taminD1 calciumversus placebo), whichmakes it hard to assess the effects ofvitamin D alone. Further, increased car-diovascular risk with calcium as well asvitamin D is suspected; this has beenunder debate and has recently been chal-lenged (11–14).Recently, the results of two large

trialsdVitamin D and Type 2 Diabetes(D2d) (15) and Diabetes Prevention withactive Vitamin D (DPVD) (16)dweremadeavailable. Both trials individually showedthat the risk of new-onset T2DM wastrending lower in the vitamin D groupthan in theplacebogroup,but thefindingswere not statistically significant. Wesought to determine whether an effectcould be observed when pooled acrosstrials. Therefore, we conducted a meta-analysis toevaluate theeffectof vitaminDsupplementation on the risk of T2DMamong people with prediabetes.

METHODS

Protocol and GuidanceThis study was performed in accordancewith Preferred Reporting Items for Sys-tematic Reviews and Meta-Analyses(PRISMA) (17). The protocol of thisstudy was registered with PROSPERO(reg. no. CRD42019139881).

Selection Criteria

Inclusion Criteria

1) Population: adults (age$18years)withprediabetes. Prediabetes was definedas impaired fasting glucose (WorldHealth Organization [WHO] criteria [18],

6.1–6.9 mmol/L, or the American Di-abetes Association definition [19],5.6–6.9 mmol/L), impaired glucosetolerance (2-h plasma glucose 7.8–11.0 mmol/L during an oral glucosetolerance test) (18), or raised glycosy-lated hemoglobin (HbA1c) (AmericanDiabetes Association criteria [19], 39–47 mmol/mol, or National Institute forHealth and Care Excellence criteria [20],42–47 mmol/mol).

2) Intervention: vitamin D supplements,regardless of the type, dose, duration,or route of administration.

3) Comparison intervention: placebo. Tri-als with multiple interventions (e.g., co-administered vitamin D and calcium)were eligible if the study groups differedonly by the use of vitamin D.

4) Outcome(s): the primary outcome wasnew-onset T2DM. Secondary outcomewas reversion of prediabetes to normal.

5) Randomized controlled trial (includ-ing both quasi-randomized and cluster-randomized trials).

Exclusion Criteria

1) Case reports, case series, and obser-vational studies.

2) Trials of participants with diabetes.3) Trials with duration ,6 months.

Data Sources and SearchesSearches were conducted by an experi-enced research librarian (P.X.) in the data-basesMEDLINE (Ovid), Embase (Ovid), andthe Cochrane Central Register of Con-trolled Trials (CENTRAL) from inception to5 July 2019, without language restrictions.We also searched the WHO InternationalClinical Trials Registry Platform to identifyongoing or unpublished potentially eli-gible trials. To maximize our search forrelevant articles, we manually reviewedthe reference lists of identified trials andsystematic reviews. The search strategy ispresented in Supplementary Table 1.

Study SelectionAfter removal of duplicates, two reviewers(Y.Z. and H.T.) independently screened alltitles and abstracts for potential relevance,with full texts obtained for those deemedrelevant. Disagreements were resolved byconsensus or with input from a third in-dependent reviewer (F.F.).

Data Extraction and QualityAssessment

Data Collection Process

Two reviewers (Y.Z. and H.T.) indepen-dently extracted data from the included

trials using a standard data extractionform (Supplementary Table 2). We con-tacted the corresponding authors formiss-ing or unreported data. Disagreementswere resolved by consensus or with inputfrom a third independent reviewer (F.F.).

Assessment of Risk of Bias and Quality of

Evidence

Two reviewers (Y.Z. and H.T.) indepen-dently performed risk-of-bias assess-ment using the Cochrane Collaborationrisk of bias (RoB) tool across seven do-mains: random sequence generation;allocation concealment; blindingof studyparticipants, health care providers, andoutcome assessors; incomplete outcomedata; and other potential sources of bias(21). Each domain was assessed as low,unclear, or high risk of bias. The trialswere judged as low risk of bias when alldomains were judged as low risk of bias.Conversely, trialswere judgedashigh riskof bias when one or more domains werejudged as unclear or high risk of bias. Ifdata of risk estimates were unavailablefrom published reports, we collectedrelevant data by protocol, correspond-ing with the authors, and abstractingfrom previous systematic reviews. Tworeviewers (Y.Z. and Y.F.) independentlyrated their confidence in the effect esti-mates of each outcome using the GRADE(Grading of Recommendations, Assess-ment, Development and Evaluation)approach (22).

Data Synthesis and AnalysisWe performed statistical analyses usingReview Manager (5.3.3; the CochraneCollaboration) and the meta package inR (version 3.4.3; R Project for StatisticalComputing). Continuous variables werepooled using the inverse variance random-effects model and were presented asmean differences with 95% CIs. Di-chotomous variables were pooled usingthe Mantel-Haenszel method and pre-sented as risk ratios (RRs) with 95% CI.We conducted a secondary analysis forthe primary outcomes of interest usinghazard ratios (HRs) and 95% CI in orderto account for the variation in follow-upbetween the included trials and to as-sess the impact of censoring. We consid-ered a P value of ,0.05 statisticallysignificant. We assessed heterogene-ity using the I2 test (23). We used afixed-effects weighted model to cal-culate the pooled estimates exceptwhere I2 . 30%, in which case a

care.diabetesjournals.org Zhang and Associates 1651

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/43/7/1650/630440/dc191708.pdf by guest on 04 January 2022

random-effects weighted model wasused. The possibility of small study effectswas assessed qualitatively by a visualestimate of funnel plot and quantita-tively by calculation of the Egger test,Begg test, and Harbord test (24).

Subgroup Analysis

Several a priori determined subgroupanalyses were performed to test inter-actions according to dose ($2,000 IU/day and,2,000 IU/day), type of vitaminD (vitamin D2 and vitamin D3), timing oftreatment (daily and intermittently),meanBMI([weight inkilogramsdividedbythe square of the height in meters] $30kg/m2 and ,30 kg/m2), mean baseline25(OH)D ($50 nmol/L and,50 nmol/L),length of maximum follow-up ($3 yearsand,3 years), intervention (vitamin D vs.placebo and adjunctive vitamin D to cal-cium vs. calcium alone), and latitude($37° and ,37°). We conducted posthoc subgroup analyses based on meanbaseline 25(OH)D ($30 nmol/L and ,30nmol/L) and achieved 25(OH)D in thevitamin D group ($75 nmol/L and ,75nmol/L).

Sensitivity Analyses

Sensitivity analyseswere conductedby1)excluding studies with a high risk of bias,2) excluding small trials (number of par-ticipants ,200), 3) excluding quasi-ran-domized or cluster-randomized trials, 4)excluding trials with follow-up ,1 year,5) using random-effects or fixed-effectsmodel, 6) excluding trials reported as anabstract, and 7) excluding the largesttrial.

Data and Resource Availability.Data are available from the correspond-ing author.

RESULTS

Eligible Studies and StudyCharacteristicsThe PRISMA flow diagram of the meta-analysis is shown in Fig. 1. Our systematicelectronic literature search initially iden-tified 4,265 studies. After application ofexclusion criteria, eight trials (15,16,25–30)were deemed eligible and included in themeta-analysis. The excluded trials and rea-sons forexclusionare listed inSupplementaryTable 3.The summary characteristics of the in-

cluded trials are shown in Table 1, and theinclusion criteria of those trials are shownin Supplementary Table 4. All studieswere

published from 2013 to 2019. A total of4,896 participants were included, withsample sizes ranging from 147 to 2,423,and two trials (15,16) included .1,000participants. Five trials recruited partic-ipants with prediabetes (15,16,28–32),two trials recruited participants withprediabetes and vitamin D deficiency(25,26), and one trial recruited partic-ipants with prediabetes, vitamin D de-ficiency, and obesity (27). The durationof follow-up of the included trials rangedfrom 6 months to 5 years.

The riskofbias is shown inSupplementaryFigs. 1 and 2. Three trials (15,16,29) had alow risk of bias, one (25) had unclear risk,and trials (26–28,30)hadahigh risk.Mosttrials thatwereassignedahigh riskofbiashad insufficient or nonblinding of par-ticipants. The quality of evidence wasrated following theGRADEapproach. Thequality of the primary outcome is mod-erate (Supplementary Table 5).

New-Onset T2DMAll eight trials (15,16,25–30) measuredand reported the development of new-onset diabetes. In total, 1,022 (20.9%) of4,896 trial participants were diagnosedwith new-onset diabetes during the tri-al. Combining data from all eight trials(15,16,25–30) reporting a RR, we foundthat vitamin D supplementation reducedthe incidence of new-onset diabetes by11% (RR 0.89 [95% CI 0.80–0.99]; I25 0%[Fig. 2A]). When combining data from thethreetrials (15,16,29) thatreportedanHR,we also found a similar decrease in new-onset diabetes in those with vitamin Dsupplementation compared with controlsubjects (HR 0.88 [95%CI 0.78–0.99]; I250% [Fig. 2B]).

Funnel plot analysis showed no asym-metry (Supplementary Fig. 3). The Eggertest (P5 0.11), Begg test (P5 0.72), andHarbord test (P 5 0.16) did not detectany significant small study effects.

Foreachofour sensitivity analyses, thepointestimatesformostoutcomeschangedminimally (Table 2). Five of 16 sensitivityanalyses changed in statistical significance.Whenwe excluded studies with a high riskof bias, the pooledHR showed a significantreduction in diabetes (HR 0.88 [95% CI0.78–0.99]), although the pooled RR didnot (RR 0.91 [95% CI 0.81–1.01]). Withsmall trials excluded, the pooled HR wassignificant (HR 0.88 [95% CI 0.78–0.99]),but thepooledRRwasnot (RR0.90 [95%CI0.81–1.01]). With exclusion of the largest

trial, both the pooled HR (0.88 [95% CI0.74–1.05]) and the pooled RR (0.87 [95%CI 0.73–1.02]) were nonsignificant. Whenexcluding we excluded trials only availableas an abstract, the pooled RR showed asignificant reduction in diabetes (RR 0.89[95%CI0.80–1.00],P50.04), althoughthepooled HR did not (HR 0.89 [95% CI 0.77–1.02]). The results of other sensitivity anal-yses were robust (Table 2).

Subgroup analyses found that vitaminD did not lower the risk of new-onsetdiabetes among obese patients (meanBMI $30 kg/m2, RR 0.95 [95% CI 0.84–1.08]; I2 5 0%). However, vitamin D re-duced the risk of new-onset diabetes innonobese patients by 27% (mean BMI,30 kg/m2, RR 0.73 [95% CI 0.57–0.92];I2 5 4%; Pinteraction 5 0.048 [Table 3 andSupplementary Fig. 4]). However, we didnot find any interaction accordingto mean baseline 25(OH)D, achieved25(OH)D in vitaminDgroup, dose, lengthof follow-up, intervention, or latitude.Similarly, meta-regression examining theeffect of dose and duration of vitamin Don diabetes risk did not reveal a statis-tically significant effect (SupplementaryFigs. 5 and 6).

Reversion of Prediabetes to NormalFive trials (25–27,29,30) totaling 1,080participants reported the rate of rever-sion of prediabetes to normal. The re-versionof prediabetes to normoglycemiaoccurred in 116 of 548 (21.2%) partic-ipants in the vitamin D group and 75 of532 (14.1%) in the control group. Com-paring the two groups, we find that therate of reversion from prediabetes tonormoglycemiawas significantly increasedby vitamin D supplementation (RR 1.48[95% CI 1.14–1.92]; I25 0% [Fig. 3]). Theresults were consistent among differentsubgroups, and there was no statisticalevidence of interaction (SupplementaryTable6).However, theRR for reversionofprediabetes to normal in obese patients(1.93 [95% CI 1.00–3.71]) was trendinghigher than that in nonobese patients(1.41 [95% CI 1.06–1.86]), although theinteraction was not statistically signifi-cant (P 5 0.38).

DISCUSSION

In thismeta-analysis of eight randomizedcontrolled trials involving patients withprediabetes, vitamin D supplementationreduced the risk of new-onset T2DM.Similarly, our meta-analysis suggested

1652 Vitamin D and T2DM Prevention in Prediabetes Diabetes Care Volume 43, July 2020

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/43/7/1650/630440/dc191708.pdf by guest on 04 January 2022

thatvitaminDsupplementationmay leadto increased reversion of prediabetes tonormal.Two major mechanisms potentially ex-

plainhowvitaminDsupplementationmayreduce T2DM risk. First, vitamin D mod-ulates insulin synthesis and secretion. Thepancreatic islet cells contain all compo-nents of the vitamin D endocrine system,including the vitamin D receptor (33), 1a-hydroxylase (34), and vitamin D–bindingprotein (35). In animal studies, vitamin Ddeficiency is associated with decreasedinsulin synthesis and secretion (36),whereas vitamin D supplementation re-stored insulin secretion (37). Moreover,vitamin D modulates the local pancreaticislet renin-angiotensin system while im-proving isletb-cell secretory function(38).Second, in peripheral insulin-target cells,vitaminDmay reduce insulin resistance inseveral ways, including presence of thevitamin D receptor in adipocytes (39),muscle (40), and hepatocytes (41); fa-cilitation of the expression of insulin re-ceptor and insulin responsiveness forglucose transport (42); and indirectly via

the regulation of calcium metabolism,which is essential for insulin-mediatedintracellular processes (40). These mech-anisms support the continued interestamong clinical researchers in using vi-taminDsupplementationtoreduceT2DMrisk.

Comparison With Other StudiesPublished systematic reviews have alsoevaluated the effect of vitamin D sup-plementation on glycemic control anddiabetes prevention in people with andpeople without prediabetes (6–10). Char-acteristics and included trials of thosereviewsarepresented inSupplementaryTables 7 and 8. Those reviews mostlyfound that vitamin D supplementationwas not associated with a lower inci-denceof diabetes (6–8). In 2014, ameta-analysis of four trials by Seida et al. (6)found that vitamin D3 supplementationhad no preventive effect on diabetes inpatients with prediabetes (odds ratio1.02 [95% CI 0.94–1.10]). For popula-tions without diabetes, a review in 2018by Tang et al. (8) found that vitamin D

supplementation did not affect the risk ofT2DM (RR 1.01 [95%CI 0.93–1.08]). In thesame year, an additional systematic re-viewbyHeetal. (7) suggestedthatvitaminD supplementation did not reduce theincidence of T2DM in the adults withoutdiabetes (RR 0.86 [95%CI 0.74–1.01]), butthe incidence of T2DMwas decreased in asubgroup analysis of people with predi-abetes (RR 0.84 [95% CI 0.70–1.00]; P 50.047). However, that meta-analysis in-correctly included two reports (29,43) ofthe same trial with different follow-updurations, and that trial contributedthe most participants to that analysis.With exclusion of that one study alone,the subgroup analysis was no longersignificant.

The major difference between thecurrent study and previous reviews wasthe inclusion of three large, relativelylong-term (3–5 years) trials, accountingfor 78.5% (3,762 of 4,788) of the totalnumber of participants in this analysis.The data from these trials were notpreviously available and provide improvedstatistical power concerning the effectsof vitamin D supplementation on T2DMprevention. Further, the two largest trials(D2d and DDVP) (15,16) had nonsignif-icant findings on their own, and it wasstated that D2d was powered to detect25% risk reduction, but the true effectwas lower (15).

Effects of BMI in Modulating theBenefits of Vitamin DAn interesting finding of the subgroupanalysis is that the benefit of vitamin Dsupplementation on prevention of dia-beteswasonly seen innonobesepatientsand not in obese patients. Several the-ories may explain the observed finding.First, vitamin D is fat soluble and is moreeasily sequestered into fat cellsandstoreduntil needed for furthermetabolism. Stud-ies have shown that obese patients needhigher loading doses of vitamin D thannormal-weightpatients toachieveasimilarincrease in serum 25(OH)D concentration(44). Second, there is increasing evidencethat obesity influences 25(OH)D metabo-lism (45). Obesity decreases hepatic 25-hydroxylaseactivity,which in turnreducesserum 25(OH)D (46).

However, without individual patientdata, subgroup analysis ofmean baselineBMI is not reliable. Trials with mean base-line BMI ,30 kg/m2 may include partic-ipants with BMI $30 kg/m2 in both

Figure 1—PRISMA flow diagram.

care.diabetesjournals.org Zhang and Associates 1653

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/43/7/1650/630440/dc191708.pdf by guest on 04 January 2022

vitaminD–treatedand–untreated groups.Likewise, the converse (i.e., trials withmean baseline BMI $30 kg/m2 mayinclude participants with BMI ,30 kg/m2) could occur. Moreover, the findingfrom the subgroup analysis should beinterpreted with caution because theanalysis was influenced by one largemultisite study, D2d.

Strengths and LimitationsThe strengths of this review included acomprehensive search for evidence, us-ing a priori research protocol, duplicateassessments of eligibility, risk of bias,and data abstraction. The study includeda rigorous assessment of the quality ofevidence and of the credibility of sub-group analyses.Several limitations must be consid-

ered. First, some trials included in this

systematic review differed in their def-inition of prediabetes and new-onsetdiabetes. For example, some trials de-fined impaired plasma glucose, in ac-cordance withWHO criteria, as a fastingblood glucose of 6.1–6.9mmol/L; othersfollowed the American Diabetes Asso-ciation guideline, which recommended acutoff point of 5.6–6.9 mmol/L.

Second, the analysis included only trial-level data. Only themain trial results wereconsidered. Individual patient data werenot available. Individual patient data helpdetermine, for example, whether the ben-efit of vitamin D is restricted to obesesubjects. Our subgroup analysis was basedon mean baseline BMI, and its reliabilitywould be improvedwith individual patientdata.

Third, while included trials focused onshort-term or intermediate outcomes,

the prodromal period of prediabetes thatproceeds toward T2DM may be as longas a decade (47). Participants enrolled inshort-termtrials are likely tobeat varyingstages of b-cell damage for the durationof follow-up, and therefore participantsmay vary in their degree of response toattempts at b-cell rescue. This inevitablyreduces the effect seen in a trial setting,which implies that any reductions inT2DM risk with vitamin D supplemen-tation are likely to be larger in real lifethan in the context of a trial. Thus,longer-term follow-up is required to ex-amine whether the beneficial effect ofvitamin D on prevention of T2DM wassustained, or even increased, with reple-tion over extended periods.

Fourth, the participants of the includedtrials were of varying ethnicity, were fromdifferent countries, and potentially had

Table 1—Summary characteristics of included studies

First author,year(reference no.) Country

Studypopulation

No. ofpatients

%female

Meanage

(years)

MeanBMI

(kg/m2) Treatment Control Duration

Baseline25(OH)D

forvitaminD/controlgroups(nmol/L)

Achieved25(OH)Dfor vitaminD/controlgroups(nmol/L)

Davidson,2013 (25)

U.S. Prediabetesand vitamin D

deficient

117 67.9 52.4 32.5 Vitamin D3 88,865IU/week

Placebo 1 year 52.5/52.5 162.5/50.5

Dutta,2014 (26)

India Prediabetesand vitamin D

deficient

170 59.2 47.9 26.6 Vitamin D3 60,000IU/week for8 weeks, then60,000 IU/

month 1 calcium500 mg/day

Calcium500

mg/day

1 year 42.5/45 88.4/43.8

Barengolts,2015 (27)

U.S. Prediabetes,vitamin Ddeficiency,

andoverweight

173 0 59.0 31.9 Vitamin D2 50,000IU/week

Placebo 1 year 36.7/35 120/50

Kuchay,2015 (28)

India Prediabetes 147 NA 48.1 25.5 Vitamin D3

60,000 IU/week for 4 weeks,then 60,000 IU/

month

Placebo 1 year 49.4/47.1 108/55.9

Jorde,2016 (29)

Norway Prediabetes 511 38.6 62 30.0 Vitamin D3 20,000IU/week

Placebo 5 years 59.9/61.1 122.3/66.7

Kawahara,2018 (16)

Japan Prediabetes 1,256 NA NA NA Eldecalcitol0.75 mg/day

Placebo 3 years NA NA

Niroomand,2019 (30)

Iran Prediabetes 162 76.5 46.5 31.5 Vitamin D3 50,000IU/week for 3months, then

50,000 IU/month

Placebo 6 months 30.7/31.7 89.9/39.9

Pittas,2019 (15)

U.S. Prediabetes 2,423 44.8 60 32.1 Vitamin D3 4,000IU/day

Placebo 5 years 69.2/70.5 135.8/72.0

Data are presented as mean unless otherwise indicated. To convert vitamin D IU to mg, multiply by 0.0025. NA, not available.

1654 Vitamin D and T2DM Prevention in Prediabetes Diabetes Care Volume 43, July 2020

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/43/7/1650/630440/dc191708.pdf by guest on 04 January 2022

different dietary habits; such charac-teristics may modulate the effects ofvitamin D supplementation.Fifth, there are various methods for

assessing vitamin D status serum via25(OH)D serum levels, including protein-binding assays and liquid chromatography–tandem mass spectrometry. Among theeight included trials, the method of

measuring 25(OH)D levels varied acrosssome trials and was not reported inseveral.

ImplicationsIn 2019, the American Diabetes Associ-ation released a position statement ofnutrition therapy recommendations forthemanagement of adultswith diabetes.

The statement does not recommend theroutine use of vitamin D ormultivitaminsfor improving glycemia in people withdiabetes or prediabetes due to a lack ofevidence (48). However, the AmericanDiabetes Association recommends in-tensive lifestyle intervention and met-formin for diabetes prevention. In theDiabetes Prevention Program, lifestyle

Figure 2—Forest plot of new-onset T2DM for all trials evaluating vitamin D supplementation. A: Pooled RR. B: Pooled HR. df, degrees of freedom; IV,inverse variance; M-H, Mantel-Haenszel.

Table 2—Sensitivity analyses of the primary outcome

Sensitivityanalyses

No. of excludedtrials (reference

nos.)

No. of includedtrials (reference

nos.) HR (95% CI) P

No. of includedtrials (reference

nos.) RR (95% CI) P

Overall analysis 0 3 (15,16,29) 0.88 (0.78, 0.99) 0.04 8 (15,16,25–30) 0.89 (0.80, 0.99) 0.03

Excluding studies withhigh risk of bias 5 (25–28,30) 3 (15,16,29) 0.88 (0.78, 0.99) 0.04 3 (15,16,29) 0.91 (0.81, 1.01) 0.08

Excluding small trial(participants ,200) 5 (25–28,30) 3 (15,16,29) 0.88 (0.78, 0.99) 0.04 3 (15,16,29) 0.91 (0.81, 1.01) 0.08

Excluding quasi/cluster-randomized trials 0 3 (15,16,29) 0.88 (0.78, 0.99) 0.04 8 (15,16,25–30) 0.89 (0.80, 0.99) 0.03

Excluding trials withfollow-up ,1 year 1 (30) 3 (15,16,29) 0.88 (0.78, 0.99) 0.04 7 (15,16,25–29) 0.90 (0.81, 1.00) 0.04

Using random-effectsmodels 0 3 (15,16,29) 0.88 (0.78, 0.99) 0.04 8 (15,16,25–30) 0.90 (0.81, 1.00) 0.04

Excluding trials reportedas abstract 1 (16) 2 (15,29) 0.89 (0.77, 1.02) 0.09 7 (15,25–30) 0.89 (0.80, 1.00) 0.04

Excluding the largest trial 1 (15) 2 (16,29) 0.88 (0.74, 1.05) 0.17 7 (16,25–30) 0.87 (0.73, 1.02) 0.09

care.diabetesjournals.org Zhang and Associates 1655

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/43/7/1650/630440/dc191708.pdf by guest on 04 January 2022

intervention reduced the risk of pro-gression to diabetes by 58% comparedwith placebo (49). Our meta-analysisshowed that in patients with predia-betes, vitamin D supplementation re-duced the risk of T2DMby 11%, and thisrisk can be further reduced by reversingprediabetes to the normoglycemic stateby 48%. Although the magnitude of the

effect of vitamin D was smaller than thatof lifestyle intervention, achieving andmaintaining a lifestyle-modificationpro-gram (defined as 150 min of physicalactivity per week and.7% weight loss)is challenging; vitamin D is a safe, eco-nomical, and widely available nutrient.Thus, vitamin D may play a role in theprevention of diabetes in these who

cannot sustain an intensive lifestyle in-tervention. Moreover, it remains to beseen whether adjunctive vitamin D inpeople with prediabetes receiving in-tensive lifestyle intervention mightlower the incidence of T2DM furtherthan lifestyle intervention alone. Ad-ditionally, others have observed in-creased insulin resistance during the

Table 3—Subgroup analysis of the effect of vitamin D on new-onset diabetes

No. of trials No. of patients I2 RR (95% CI) Pinteraction

Overall 8 4,896 0% 0.89 (0.80, 0.99)

Baseline 25(OH)D (nmol/L)$50 3 2,517 0% 0.90 (0.80, 1.02) 0.79,50 5 1,122 26% 0.87 (0.68, 1.11)

Baseline 25(OH)D (nmol/L)$30 7 3,640 0% 0.89 (0.80, 1.00) NA,30 0 0 NA NA

Achieved 25(OH)D in vitamin D group (nmol/L)$75 7 4,734 0% 0.90 (0.81, 1.00) NA,75 0 0 NA NA

Type of vitamin DVitamin D3 7 4,723 0% 0.89 (0.80, 0.99) 0.81Vitamin D2 1 173 NA 0.99 (0.41, 2.37)

Daily dose equivalent (IU)$2,000 8 4,896 0% 0.89 (0.80, 0.99) NA,2,000 0 0 NA NA

Mean BMI (kg/m2)

$30 5 2,514 0% 0.95 (0.84, 1.08) 0.048,30 3 1,126 4% 0.73 (0.57, 0.92)

Timing of treatmentDaily 2 3,679 0% 0.90 (0.80, 1.03) 0.70Intermittently 6 1,217 8% 0.86 (0.66, 1.10)Intervention

Vitamin D vs. placebo 7 4,771 0% 0.90 (0.81, 1.01) 0.07

Vitamin D 1 calcium vs. calcium 1 125 NA 0.39 (0.16, 0.95)

Latitude$37° 3 3,557 0% 0.89 (0.79, 1.00) 0.85,37° 6 1,339 15% 0.86 (0.63, 1.18)

Follow-up$3 years 3 4,190 0% 0.91 (0.81, 1.01) 0.28,3 years 5 706 0% 0.70 (0.45, 1.10)

To convert Vitamin D IU to mg, multiply by 0.0025. NA, not available.

Figure 3—Forest plot of reversion of prediabetes to normal for all trials evaluating vitamin D supplementation. M-H, Mantel-Haenszel.

1656 Vitamin D and T2DM Prevention in Prediabetes Diabetes Care Volume 43, July 2020

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/43/7/1650/630440/dc191708.pdf by guest on 04 January 2022

long run-in period for T2DM develop-ment, and that insulin resistance canbe reduced by vitamin D supplementa-tion when 25(OH)D values are raised tosufficient levels, at least in persons withvitamin D deficiency (45,50).Nonetheless, these findings should be

interpretedwith cautionbecause theover-all RR was relatively small and statisticallymarginal. In particular, sensitivity analysesshowed the results were largely driven bythe largest trial (D2d).

ConclusionResults from this meta-analysis show thatvitamin D supplementation reduces therisk of T2DM in participants with pre-diabetes. Reversion of prediabetes tonormoglycemia was significantly increasedby vitamin D supplementation. The ben-efit of thepreventionof T2DMappears tobe confined to nonobese subjects. In-dividual participant data meta-analysesare needed to confirm the overall resultand identify subgroups that benefit themost from supplementation.

Funding. This work is supported by projects ofthe National Natural Science Foundation of China(81100925and81472361) andbytheNationalKeyR&D Program of China (2018YFA 0108604).The funders of the study had no role in study

design, data collection, data analysis, data in-terpretation, or writing of the manuscript.Duality of Interest. No potential conflicts ofinterest relevant to this article were reported.Author Contributions. F.F. and Y.Z. contrib-uted to study conceptanddesign. P.X. performedthe literature search. Y.Z. and H.T. selected thestudies. Y.Z. and H.T. performed data extractionfrom relevant studies. J.T., J.L., and Y.Z. preparedinitial figures and data analysis. Y.Z. wrote thefirst draft of the manuscript. Y.Z., H.T., J.T., J.L.,W.C., Y.H., Y.F., L.D.L., P.X., D.J., and F.F. criticallyrevised successive drafts of the manuscript andapproved the final version. F.F. has affirmed thatthis manuscript is an honest, accurate, andtransparent account of the study being re-ported; that no important aspects of the studyhave been omitted; and that any discrepanciesfrom the study as planned (and, if relevant,registered) have been explained. F.F. is theguarantor of this work and, as such, had fullaccess to all the data in the study and takesresponsibility for the integrity of the data andthe accuracy of the data analysis.

References1. Pittas AG, Lau J, Hu FB, Dawson-Hughes B. Theroleof vitaminDandcalcium in type2diabetes.Asystematic review and meta-analysis. J Clin En-docrinol Metab 2007;92:2017–20292. Tai K, Need AG, Horowitz M, Chapman IM.Glucose tolerance and vitamin D: effects of

treating vitamin D deficiency. Nutrition 2008;24:950–9563. KobzaVM,Fleet JC,ZhouJ,etal.VitaminDstatusand resistance exercise training independentlyaffect glucose tolerance in older adults. Nutr Res2013;33:349–3574. Kayaniyil S, Vieth R, Retnakaran R, et al. As-sociation of vitamin D with insulin resistance andbeta-cell dysfunction in subjects at risk for type 2diabetes. Diabetes Care 2010;33:1379–13815. Zeitz U, Weber K, Soegiarto DW, Wolf E,Balling R, Erben RG. Impaired insulin secretorycapacity in mice lacking a functional vitamin Dreceptor. FASEB J 2003;17:509–5116. Seida JC, Mitri J, Colmers IN, et al. Clinicalreview: effect of vitamin D3 supplementationon improving glucose homeostasis and pre-venting diabetes: a systematic review andmeta-analysis. J Clin Endocrinol Metab 2014;99:3551–35607. HeS, YuS, ZhouZ,WangC,WuY, LiW.Effectofvitamin D supplementation on fasting plasmaglucose, insulin resistance and prevention oftype 2 diabetes mellitus in non-diabetics: a sys-tematic review and meta-analysis. Biomed Rep2018;8:475–4848. TangH, Li D, Li Y, Zhang X, Song Y, Li X. Effectsof vitamin D supplementation on glucose andinsulin homeostasis and incident diabetesamong nondiabetic adults: a meta-analysisof randomized controlled trials. Int J Endocri-nol 2018;2018:79087649. Mirhosseini N, Vatanparast H, Mazidi M,Kimball SM. Vitamin D supplementation, glycemiccontrol, and insulin resistance in prediabetics:a meta-analysis. J Endocr Soc 2018;2:687–70910. Poolsup N, Suksomboon N, Plordplong N.Effect of vitamin D supplementation on insulinresistance and glycaemic control in prediabetes:a systematic review and meta-analysis. DiabetMed 2016;33:290–29911. Anderson JL, May HT, Horne BD, et al.;Intermountain Heart Collaborative (IHC) StudyGroup. Relation of vitamin D deficiency to car-diovascular risk factors, disease status, and in-cident events in a general healthcare population.Am J Cardiol 2010;106:963–96812. Bolland MJ, Avenell A, Baron JA, et al. Effectof calcium supplements on risk of myocardialinfarction and cardiovascular events: meta-anal-ysis. BMJ 2010;341:c369113. Lewis JR, Zhu K, Prince RL. Adverse eventsfrom calcium supplementation: relationship toerrors in myocardial infarction self-reporting inrandomized controlled trials of calcium supple-mentation. J Bone Miner Res 2012;27:719–72214. Bolland MJ, Grey A, Avenell A, Gamble GD,Reid IR. Calcium supplements with or withoutvitamin D and risk of cardiovascular events:reanalysis of theWomen’sHealth Initiative limitedaccess dataset and meta-analysis. BMJ 2011;342:d204015. Pittas AG, Dawson-Hughes B, Sheehan P,et al.; D2d Research Group. Vitamin D supple-mentation and prevention of type 2 diabetes. NEngl J Med 2019381:520–53016. Kawahara T. Eldecalcitol, a vitamin D analog,for diabetes prevention in impaired glucosetolerance (DPVD study) (Abstract). Diabetes 2018;67(Suppl. 1):120-LB

17. Liberati A, Altman DG, Tetzlaff J, et al. ThePRISMA statement for reporting systematicreviews and meta-analyses of studies thatevaluate health care interventions: explanationand elaboration. Ann Intern Med 2009;151:W65–W9418. WHO/IDF Consultation. Definition and di-agnosis of diabetes and intermediate hypergly-caemia [Internet], 2006. Available from http://www.who.int/diabetes/publications/Definition%20and%20diagnosis%20of%20diabetes_new.pdf. Accessed 1 March 201519. Expert Committee on the Diagnosis and Clas-sification of DiabetesMellitus. Report of the ExpertCommittee on the Diagnosis and Classification ofDiabetesMellitus.DiabetesCare2003;26(Suppl. 1):S5–S2020. ChattertonH, Younger T, Fischer A, Khunti K;Programme Development Group. Risk identifi-cation and interventions to prevent type 2 di-abetes in adults at high risk: summary of NICEguidance. BMJ 2012;345:e462421. Higgins JP, Altman DG, Gotzsche PC, et al.;Cochrane Statistical Methods Group. The Co-chrane Collaboration’s tool for assessing risk ofbias in randomised trials. BMJ 2011;343:d592822. Guyatt GH, OxmanAD, Vist GE, et al.; GRADEWorking Group. GRADE: an emerging consensuson rating quality of evidence and strength ofrecommendations. BMJ 2008;336:924–92623. Higgins JP, Thompson SG. Quantifying het-erogeneity in ameta-analysis. StatMed 2002;21:1539–155824. Egger M, Davey Smith G, Schneider M,Minder C. Bias in meta-analysis detected by asimple, graphical test. BMJ 1997;315:629–63425. DavidsonMB,DuranP, LeeML, FriedmanTC.High-dose vitamin D supplementation in peoplewith prediabetes and hypovitaminosis D. Diabe-tes Care 2013;36:260–26626. Dutta D, Mondal SA, Choudhuri S, et al.Vitamin-D supplementation in prediabetes re-duced progression to type 2 diabetes and wasassociated with decreased insulin resistance andsystemic inflammation: an open label random-ized prospective study from Eastern India. Di-abetes Res Clin Pract 2014;103:e18–e2327. Barengolts E, Manickam B, Eisenberg Y,Akbar A, Kukreja S, Ciubotaru I. Effect of high-dose vitamin D repletion on glycemic control inAfrican-American males with prediabetes andhypovitaminosis D. Endocr Pract 2015;21:604–61228. Kuchay MS, Laway BA, Bashir MI, Wani AI,Misgar RA, Shah ZA. Effect of Vitamin D supple-mentation on glycemic parameters and progres-sion of prediabetes to diabetes: a 1-year, open-label randomizedstudy. Indian JEndocrinolMetab2015;19:387–39229. JordeR, Sollid ST, Svartberg J, et al. VitaminD20,000 IU per week for five years does notprevent progression from prediabetes to diabe-tes. J Clin Endocrinol Metab 2016;101:1647–165530. Niroomand M, Fotouhi A, Irannejad N,Hosseinpanah F. Does high-dose vitamin Dsupplementation impact insulin resistance andrisk of development of diabetes in patientswith pre-diabetes? A double-blind randomizedclinical trial. Diabetes Res Clin Pract 2019;148:1–931. Iraj B, Aminorroaya A, Amini M. Does theintramuscular injection of vitamin D increase

care.diabetesjournals.org Zhang and Associates 1657

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/43/7/1650/630440/dc191708.pdf by guest on 04 January 2022

insulin resistance? J Res Pharm Pract 2012;1:60–6532. Forouhi NG, Menon RK, Sharp SJ, et al.Effects of vitamin D2 or D3 supplementation onglycaemic control and cardiometabolic risk amongpeople at risk of type 2 diabetes: results of arandomized double-blind placebo-controlled trial.Diabetes Obes Metab 2016;18:392–40033. Ishida H, Norman AW. Demonstration of ahigh affinity receptor for 1,25-dihydroxyvitaminD3 in rat pancreas. Mol Cell Endocrinol 1988;60:109–11734. Bland R, Markovic D, Hills CE, et al. Expres-sion of 25-hydroxyvitamin D3-1alpha-hydroxy-lase in pancreatic islets. J Steroid Biochem MolBiol 2004;89-90:121–12535. WootenMW,NelAE,Goldschmidt-ClermontPJ, Galbraith RM, Wrenn RW. Identification of amajor endogenous substrate for phospholipid/Ca21-dependent kinase in pancreatic acini as Gc(vitamin D-binding protein). FEBS Lett 1985;191:97–10136. Norman AW, Frankel JB, Heldt AM, GrodskyGM. Vitamin D deficiency inhibits pancreatic se-cretion of insulin. Science 1980;209:823–82537. Cade C, Norman AW. Vitamin D3 improvesimpaired glucose tolerance and insulin secretionin the vitamin D-deficient rat in vivo. Endocri-nology 1986;119:84–90

38. ChengQ, Li YC, Boucher BJ, LeungPS. Anovelrole for vitamin D: modulation of expression andfunction of the local renin-angiotensin system inmouse pancreatic islets. Diabetologia 2011;54:2077–208139. Kamei Y, Kawada T, Kazuki R, Ono T, Kato S,Sugimoto E. Vitamin D receptor gene expressionis up-regulated by 1, 25-dihydroxyvitamin D3 in3T3-L1 preadipocytes. Biochem Biophys ResCommun 1993;193:948–95540. Costa EM, Blau HM, Feldman D. 1,25-dihydroxyvitamin D3 receptors and hormonalresponses in cloned human skeletal muscle cells.Endocrinology 1986;119:2214–222041. Han S, Chiang JY. Mechanism of vitaminD receptor inhibition of cholesterol 7alpha-hydroxylase gene transcription in humanhepatocytes. Drug Metab Dispos 2009;37:469–47842. Maestro B, Campion J, Davila N, Calle C.Stimulation by 1,25-dihydroxyvitamin D3 of in-sulin receptor expression and insulin responsive-ness for glucose transport in U-937 humanpromonocytic cells. Endocr J 2000;47:383–39143. Sollid ST, Hutchinson MYS, Fuskevag OM,et al. No effect of high-dose vitamin D supple-mentation on glycemic status or cardiovascularrisk factors in subjects with prediabetes. Diabe-tes Care 2014;37:2123–2131

44. Walsh JS, Bowles S, Evans AL. Vitamin D inobesity. Curr Opin Endocrinol Diabetes Obes2017;24:389–39445. Bouillon R, Marcocci C, Carmeliet G, et al.Skeletal and extraskeletal actions of vitamin D:current evidence and outstanding questions. En-docr Rev 2019;40:1109–115146. Roizen JD, Long C, Casella A, et al. Obesitydecreases hepatic 25-hydroxylase activity caus-ing low serum 25-hydroxyvitamin D. J BoneMinerRes 2019;34:1068–107347. Sagesaka H, Sato Y, Someya Y, et al. Type2 diabetes: when does it start? J Endocr Soc2018;2:476–48448. Evert AB, Dennison M, Gardner CD, et al.Nutrition therapy for adults with diabetes orprediabetes: a consensus report. Diabetes Care2019;42:731–75449. Knowler WC, Barrett-Connor E, Fowler SE,et al.; Diabetes Prevention Program ResearchGroup. Reduction in the incidence of type 2diabetes with lifestyle intervention or metfor-min. N Engl J Med 2002;346:393–40350. vonHurstPR, StonehouseW,Coad J.VitaminD supplementation reduces insulin resistance inSouth Asian women living in New Zealand whoare insulin resistant and vitamin D deficient - arandomised, placebo-controlled trial. Br J Nutr2010;103:549–555

1658 Vitamin D and T2DM Prevention in Prediabetes Diabetes Care Volume 43, July 2020

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/43/7/1650/630440/dc191708.pdf by guest on 04 January 2022